20.03.2026Marketing for Social Impact Group at the 2026 ISMS Marketing Science Conference

The ISMS Marketing Science Conference is a premier annual event bringing together leading marketing scholars, practitioners, and policymakers who share an interest in rigorous scientific approaches to marketing problems. We look forward to contributing to the conversation and sharing our latest work with the community.

Below, you can find the abstracts of our presentations at the conference (presenters in bold):

Causal effects of targeting susceptible individuals on participation in a voluntary environmental program

Hui Zhang, Luca Lazzaro, Norina Anna Furrer, Federico Cammelli, Manuel Sebastian Mariani, Thomas Addoah, Rene Algesheimer, Rachael D. Garrett, Radu Tanase

Addressing the urgency of climate change requires large-scale adoption of new behaviors that conflict with existing social norms [Constantino et al., 2022]. Seeding is a network intervention through which a subset of the population is initially incentivized to adopt the new behavior with the goal of accelerating adoption in the entire population [Muller and Peres, 2019]. Various studies have proposed seeding central individuals in the network [e.g., Hinz et al., 2011]. An alternative approach selects “susceptible” individuals who are predisposed to adopt early [Valente, 1996]. Under network homophily, such individuals may cluster, facilitating rapid formation of a critical mass. Despite its theoretical appeal, large-scale causal evidence on this strategy remains limited. We test this approach through a field experiment in 38 villages participating in an environmental program promoting agroforestry and reducing deforestation among cocoa farmers in Côte d’Ivoire. We block-randomize the villages, stratified by geographical position, forest cover, and urban area, into two conditions: seeding susceptible individuals and random seeding (baseline). We measure the total number of participants in the program and their characteristics. To identify the seeds, we use a social sensing approach [Galesic et al., 2021] that leverages information from 15 randomly selected farmers in each village. We explain participation in the program as a function of seeding strategy and controls. Our research contributes by providing empirical causal evidence on the effectiveness of seeding susceptible individuals to accelerate diffusion.

Bridging choice experiments and innovation diffusion modeling to inform marketing interventions

Radu Tanase, René Algesheimer, Manuel Sebastian Mariani

Many managerial interventions aim to influence collective consumer behavior, such as accelerating the diffusion of new services or promoting sustainable consumption. To understand adoption processes and inform marketing activities, two research traditions have proceeded largely in parallel: Choice models and experiments focus on identifying heterogeneous individual-level adoption drivers, whereas innovation diffusion models typically study the aggregate diffusion dynamics using stylized behavioral assumptions. This disconnect limits the predictive accuracy of diffusion models and can lead to suboptimal marketing interventions (Peres and Muller, 2011, 2019). We propose an integrated framework that embeds empirically-estimated individual decision-making models into diffusion models. Specifically, we combine discrete choice modeling with complex contagion theory to estimate individual-level adoption thresholds from choice experiments. These thresholds capture the minimal level of social reinforcement needed before a consumer adopts (Centola and Macy, 2007). We validate this approach through two choice experiments in different contexts, showing that the estimated thresholds accurately predict individual adoption behavior out of sample. We then incorporate the estimated thresholds into diffusion simulations. The simulation results demonstrate that state-of-the-art seeding strategies – which rely solely on network structure – can perform substantially worse than strategies that integrate individual-level behavioral drivers and network structure. Our findings (in press at Nature Human Behaviour) highlight the importance of integrating behavioral measurement with diffusion modeling and offer a practical methodology for designing more effective interventions to stimulate large-scale adoption of products and behaviors.

Identifying central individuals without social network information

Luca Lazzaro, Mingmin Feng, Norina Furrer, Manon Delvaux, Federico Cammelli, Manuel Sebastian Mariani, Rene Algesheimer, Radu Tanase

By targeting central individuals in a social network, marketers and policymakers can design seeding interventions that promote information diffusion, product adoption, or behavior change. A crucial requirement is that decision makers possess complete social network information. However, in practice, social network information is often unavailable or prohibitively costly to collect. In this paper, we study the problem of identifying central individuals in the absence of social network information. Using a field study in a secluded cocoa-farming community in Côte d'Ivoire (N = 589), we collect complete social network data and elicit nominations of central individuals. This design allows us to compare two approaches for identifying central individuals without social network information: social sensing and one-hop strategies. Social sensing leverages individuals’ knowledge of the network by eliciting nominations of central individuals. One-hop strategies identify central individuals by sampling friends of randomly selected individuals, motivated by the friendship paradox—the fact that friends of random individuals have, on average, more connections. We find that individuals identified through social sensing have 1.5 times higher degree centrality than those identified through one-hop strategies. Yet, social sensing systematically conflates social visibility—observable individual characteristics—with degree centrality, leading to biased nominations that depend on who is visible rather than who is most central. By contrast, one-hop strategies select less central individuals on average but do not exhibit nomination bias. These findings reveal a fundamental trade-off between centrality and nomination bias, with direct implications for seeding interventions under information constraints.

Design of product discovery systems for digital retail platforms: A systematic review and research agenda

Maria Poiaganova, Manuel Sebastian Mariani, Rene Algesheimer

Digital retail platforms increasingly rely on search and recommender systems to shape what consumers see, consider, and purchase. Although a growing literature examines specific design choices, such as personalization or ranking logic, these effects are typically studied in isolation, with limited consideration for broader context in which they operate. As a result, managers risk adopting design choices that perform well in one setting but fail or backfire in another. This paper develops a unifying conceptual framework for a holistic, context-aware understanding of the effects of product discovery systems (PDS) - the integrated set of search and recommender tools governing product exposure in digital retail. Drawing on a systematic literature review of causal evidence and in-depth interviews with managers from leading e-commerce firms, we identify three core design levers: Presentation (how recommendations are communicated), Placement (how algorithms select and order products), and Portfolio (how multiple discovery systems are orchestrated within the platform), and specify key design choices within each lever. We show that no design choice is universally beneficial. Even widely praised features, such as personalization or explainability, can enhance consumer experience and platform performance in some contexts but undermine them in others when misaligned with user expertise, product characteristics, or institutional constraints. Our framework integrates these contextual boundary conditions for all the design choices, serving as a managerial decision-making paradigm for PDS design. By synthesizing these findings with insights from managerial interviews, we propose a comprehensive research agenda at the intersection of system design and consumer behavior.

Influencer Selection Based on Lifetime Value

Fei Wang, Reto Hofstetter, Manuel Sebastian Mariani

Influencer marketing has become a central pillar of firms’ marketing strategies. Existing research and practice predominantly adopt a one-off campaign perspective, relying on historical performance heuristics – such as follower counts or past engagement – to select influencers for product endorsements. However, this short-term focus overlooks the long-term value that sustained brand–influencer relationships can generate. We depart from this paradigm by introducing the Influencer Lifetime Value (ILV) framework, which conceptualizes and quantifies the discounted future value an influencer is expected to generate for a brand over the duration of a contract. Drawing inspiration from the customer lifetime value (CLV) paradigm in relationship marketing, ILV offers a forward-looking and data-efficient approach to influencer selection. The framework is flexible and can readily incorporate off-the-shelf machine-learning algorithms (e.g., Random Forest, XGBoost) to identify influencer candidates with high expected lifetime value, providing managers with a scalable and adaptable decision-support tool. Using data from three distinct influencer marketing empirical contexts, we show that commonly used heuristics lead to suboptimal influencer selection decisions and that their performance can be significantly improved by integrating machine-learning models within the proposed ILV framework. Our findings are particularly relevant for firms seeking to build long-term endorsement relationships with influencers through repeated collaborations.

Field evidence on the differential effectiveness of seeding interventions under heterogeneous individual resistance

Mingmin Feng, Rene Algesheimer, Radu Tanase

Classic diffusion research has long highlighted an "innovativeness–need" paradox: The adoption tends to concentrate initially among individuals with greater resources and lower resistance, whereas those with the greatest need may face greater resistance and delay adoption (Rogers, 1962). As a result, socially or economically vulnerable, less informed, or highly resistant individuals may remain unengaged even when overall adoption is high. Research on seeding interventions, which select a subset of individuals to initiate diffusion, emphasizes reach and diffusion speed rather than which types of individuals adopt (Hinz et al. 2011; Libai et al. 2013). Traditional hub-based seeding maximizes adoption by targeting highly connected individuals but may disproportionately activate low-threshold individuals who are already eager to adopt. Such strategies may therefore be less effective when the objective is to engage high-resistance individuals, who in many contexts need the adoption most. We run a field experiment promoting attendance at a cybersecurity training workshop, a socially beneficial behavior, among employees at a Swiss university. Engaging high-resistance individuals is particularly important, as those less attentive to cybersecurity practices may pose greater organizational risk. We randomly assign seeding interventions (e.g., hub-based, random seeding) to different individual networks and compare resistance in a pre-intervention survey with actual registration behavior across treatment and control groups. This design allows us to examine whether hub-based seeding primarily converts low-threshold individuals and how to shift the composition of adopters toward more resistant individuals. The intervention will be conducted in April–May, and results will be presented at the conference.